Posted by Marcin Makowski on October 13, 2004 at 15:58:14:
I found the following problem (bug?) in MCLR module:
I tried to calculate optimal geometry and vibrations for octatetraene molecule by CASSCF method with pi active space (8,8). I did it in C2h, Cs and C1 symmetry. In all symmetries I obtain the same (to machine precision) both final geometry and energy of the ground state. Unfortunately, when analytical frequencies are calculated by McKinley and MCLR modules the results are really not consistent. In C2h symmetry I obtain one imaginary (over 1000i) frequency for totally symmetric mode. Analytical frequencies calculations
in lower symmetries (Cs and C1) gave all real frequencies. Also numerical frequencies calculated in C1 symmetry (to obtain all the modes) are all real. I checked that this huge imaginary frequency from C2h calculations cannnot be removed by changing cutoffs and thresholds in SEWARD,RASSCF,ALASKA,SLAPAF,MCKINLEY and MCLR modules. I also checked Molcas 5.0, 5.2, 5.4 versions. All of them show the same (faulty?) behavior.
Yours,
Dr Marcin Makowski
Jagiellonian University
Faculty of Chemistry